Individualized management system and method for colorectal cancer risk prediction

Abstract

The invention provides an individualized management system and method for colorectal cancer risk prediction, and relates to the technical field of medical diagnosis and individualized medical treatment. The system comprises a multi-omics data processing module used for collecting a blood sample of a subject, extracting circulating free DNA, and sequencing cfDNA by adopting a simplified methylation sequencing technology to obtain final original data of DNA methylation characteristics; the risk prediction module is used for carrying out bioinformatics analysis on the original data of the final DNA methylation features by adopting a machine learning algorithm, extracting a methylation feature set, a tail end motif feature set and a breakpoint motif feature set, and carrying out risk prediction on the basis of the extracted feature sets; and predicting the colorectal cancer disease risk of the subject by using a pre-trained prediction model to obtain a prediction result. The problem that the risk prediction accuracy of the colorectal cancer is low is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical diagnosis and personalized medicine, and particularly to an individualized management system and method for colorectal cancer risk prediction. Background Art

[0002] Colorectal cancer is one of the malignant tumors with a relatively high incidence worldwide. Although existing screening methods such as colonoscopy and imaging examinations can effectively detect colorectal cancer, these methods are often costly and invasive, restricting their widespread application. In recent years, with the development of high-throughput sequencing technology, DNA methylation, as an important epigenetic marker, has gradually become an important tool for early cancer diagnosis. However, most existing methylation detection methods focus on the detection of single markers, lacking comprehensive analysis of multi-omics data and individualized management programs. How to efficiently and accurately use methylation data for colorectal cancer risk prediction and individualized management remains an urgent problem to be solved. Summary of the Invention

[0003] In order to overcome the deficiencies of the prior art, the purpose of the present invention is to provide an individualized management system and method for colorectal cancer risk prediction.

[0004] To achieve the above purpose, the present invention provides the following solutions:

[0005] An individualized management system for colorectal cancer risk prediction, comprising:

[0006] A multi-omics data processing module, configured to collect blood samples of a subject, extract circulating free DNA, and sequence the cfDNA using reduced representation bisulfite sequencing technology to obtain the original data of the final DNA methylation characteristics;

[0007] A risk prediction module, configured to perform bioinformatics analysis on the original data of the final DNA methylation characteristics using a machine learning algorithm, extract a methylation feature set, a terminal motif feature set, and a breakpoint motif feature set, and based on the extracted feature sets, use a pre-trained prediction model to predict the colorectal cancer risk of the subject to obtain a prediction result;

[0008] An individualized diagnosis and treatment guidance module, configured to provide an individualized colorectal cancer screening plan for the subject according to the prediction result and the clinical information of the subject;

[0009] A follow-up observation module, which conducts long-term follow-up observation on the subject, monitors the diagnosis and treatment effects of colorectal cancer, and establishes an electronic medical record database.

[0010] Preferably, the multi-omics data processing module includes:

[0011] A sample processing sub-module for collecting a peripheral blood sample of a subject and obtaining cfDNA from the peripheral blood sample of the subject;

[0012] A raw data acquisition sub-module for performing RRBS sequencing on the cfDNA to obtain initial raw sequencing data;

[0013] A data quality control sub-module for performing quality control on the initial raw sequencing data to obtain the raw data of the final DNA methylation characteristics.

[0014] Preferably, the risk prediction module includes:

[0015] A feature extraction sub-module for extracting features from the raw data of the final DNA methylation characteristics to obtain a methylation feature set, a terminal motif feature set, and a breakpoint motif feature set;

[0016] The model construction sub-module for constructing a colorectal cancer screening model set according to a machine learning algorithm;

[0017] A risk prediction sub-module for inputting the methylation feature set, the terminal motif feature set, and the breakpoint motif feature set into the colorectal cancer screening model set to obtain a prediction result.

[0018] Preferably, the colorectal cancer screening model set contains 28 basic prediction models, and each of the basic prediction models is constructed based on any one of the algorithms of support vector machine, random forest, XGBoost, and CatBoost.

[0019] Preferably, the data quality control sub-module includes:

[0020] A first quality control unit for using the FastQC tool to perform quality assessment on the initial raw sequencing data to remove low-quality bases and obtain first quality control data;

[0021] A second quality control unit for using Trimmomatic or Cutadapt to remove the adapter sequences of the first quality control data to obtain second quality control data;

[0022] A third quality control unit for using the Bsmap tool to align the second quality control data to the human reference genome hg19 to obtain third quality control data;

[0023] A screening unit for obtaining the reads ratio of the third quality control data aligned to the reference genome, determining whether the reads ratio meets the ratio threshold range, and if so, obtaining the raw data of the final DNA methylation characteristics.

[0024] Preferably, the feature extraction sub-module includes:

[0025] A first feature extraction unit, configured to calculate methylation states of multiple haploid of the original data of the final DNA methylation features to obtain a methylation feature set;

[0026] A second feature extraction unit, configured to extract a first preset number of bases at the 3'-end of reads of the original data of the final DNA methylation features and calculate distribution features to obtain an end motif feature set;

[0027] A third feature extraction unit, configured to extract a second preset number of upstream bases and downstream bases at the 3'-end of reads of the original data of the final DNA methylation features to obtain a breakpoint motif feature set.

[0028] Preferably, the types of the haploid include:

[0029] MM, MHL, Entropy, PDR, and CHALM.

[0030] An individualized management method for colorectal cancer risk prediction, comprising:

[0031] Collecting a blood sample of a subject, extracting circulating free DNA, and sequencing the cfDNA by using reduced representation bisulfite sequencing technology to obtain original data of final DNA methylation features;

[0032] Performing bioinformatics analysis on the original data of the final DNA methylation features by using a machine learning algorithm, extracting a methylation feature set, an end motif feature set, and a breakpoint motif feature set, and predicting the colorectal cancer risk of the subject by using a pre-trained prediction model based on the extracted feature sets to obtain a prediction result;

[0033] Providing a personalized colorectal cancer screening plan for the subject according to the prediction result and the clinical information of the subject;

[0034] Performing long-term follow-up observation on the subject, monitoring the diagnosis situation and treatment effect of colorectal cancer, and establishing an electronic medical record database.

[0035] The present invention discloses the following technical effects:

[0036] The present invention provides an individualized management system and method for colorectal cancer risk prediction, including: a multi-omics data processing module, which is used to collect blood samples of a subject, extract circulating free DNA, and sequence the cfDNA by using reduced representation bisulfite sequencing technology to obtain the original data of the final DNA methylation characteristics; a risk prediction module, which is used to perform bioinformatics analysis on the original data of the final DNA methylation characteristics by using a machine learning algorithm, extract a methylation feature set, a terminal motif feature set and a breakpoint motif feature set, and based on the extracted feature sets, use a pre-trained prediction model to predict the colorectal cancer risk of the subject to obtain a prediction result; a personalized diagnosis and treatment guidance module, which is used to provide a personalized colorectal cancer screening plan for the subject according to the prediction result and the clinical information of the subject; a follow-up observation module, which conducts long-term follow-up observation on the subject, monitors the diagnosis situation and treatment effect of colorectal cancer, and establishes an electronic medical record database. The present invention integrates high-throughput sequencing data of methylation and machine learning technology to achieve accurate prediction and individualized management of colorectal cancer risk, which is efficient, accurate and non-invasive, and is suitable for large-scale screening and clinical applications. The implementation manner can rely on hardware, software or a combination of both, and is realized through a dedicated logic and an instruction execution system in a memory (such as a microprocessor). The related devices and methods can be realized by computer instructions on a carrier medium (such as a disk, a CD / DVD-ROM, a programmable memory, a data carrier), or by a combination of software and firmware executed by a semiconductor and programmable hardware (such as an FPGA, a PLD), a processor; it has high specificity and sensitivity, not only accurately predicts the risk of colorectal cancer, but also provides customized screening suggestions (including imaging and pathological examinations) through a personalized diagnosis and treatment module. The follow-up observation module conducts long-term monitoring, constructs and updates an electronic medical record database to support subsequent research and clinical decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention, and for those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0038] Figure 1 It is a schematic structural diagram of an individualized management system for colorectal cancer risk prediction provided by an embodiment of the present invention;

[0039] Figure 2 It is a flowchart of an individualized management method for colorectal cancer risk prediction provided by an embodiment of the present invention.

[0040] Description of the reference numerals:

[0041] 1 - Multi - omics data processing module, 2 - Risk prediction module, 3 - Personalized diagnosis and treatment guidance module, 4 - Follow - up observation module. Detailed implementation manners

[0042] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0043] To make the above - mentioned objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0044] As Figure 1 shown, the present invention provides an individualized management system for colorectal cancer risk prediction, including:

[0045] A multi - omics data processing module 1, which is used to collect blood samples of subjects, extract circulating free DNA, and sequence the cfDNA using reduced representation bisulfite sequencing technology to obtain the original data of the final DNA methylation characteristics;

[0046] A risk prediction module 2, which is used to perform bioinformatics analysis on the original data of the final DNA methylation characteristics using machine learning algorithms, extract the methylation feature set, the terminal motif feature set, and the breakpoint motif feature set, and based on the extracted feature sets, use a pre - trained prediction model to predict the colorectal cancer risk of the subjects to obtain a prediction result;

[0047] A personalized diagnosis and treatment guidance module 3, which is used to provide a personalized colorectal cancer screening plan for the subjects according to the prediction result and the clinical information of the subjects;

[0048] A follow - up observation module 4, which conducts long - term follow - up observations on the subjects, monitors the diagnosis and treatment effects of colorectal cancer, and establishes an electronic medical record database.

[0049] Specifically, the cancer prediction results are divided into two situations: having cancer and not having cancer. When the prediction result is not having cancer, the system recommends that the subject conduct regular follow - up; when the prediction result is having cancer, it is recommended to conduct a colonoscopy review for further confirmation.

[0050] The colorectal cancer screening model in the risk prediction module 2 is trained through the following steps: taking the methylation feature set, the terminal motif feature set, and the breakpoint motif feature set as input values and inputting them into the first-layer integrated machine learning model, and then using whether having cancer as the output target of the second-layer model to perform iterative training and optimization on the model until satisfactory prediction performance is obtained.

[0051] The electronic database construction module in the follow-up observation module 4 can update the medical records of the subjects in real time, ensure the accuracy and integrity of the data, and provide strong support for subsequent research and clinical decision-making.

[0052] Furthermore, the multi-omics data processing module 1 includes:

[0053] A sample processing sub-module, which is used to collect the peripheral blood samples of the subjects and obtain cfDNA from the peripheral blood samples of the subjects;

[0054] A raw data acquisition sub-module, which is used to perform RRBS sequencing on the cfDNA to obtain initial raw sequencing data;

[0055] A data quality control sub-module, which is used to perform quality control on the initial raw sequencing data to obtain the raw data of the final DNA methylation features.

[0056] Specifically, the multi-omics data processing module 1 further includes: a sample processing sub-module: used to collect the peripheral blood samples of the subjects and obtain cfDNA from the peripheral blood samples of the subjects. A raw data acquisition sub-module, which is used to perform RRBS sequencing on the cfDNA obtained from the peripheral blood samples of the subjects to obtain raw sequencing data. A data quality control sub-module, which is used to perform quality control on the RRBS high-throughput sequencing data of the peripheral blood cfDNA of the subjects to obtain high-quality data with qualified quality control.

[0057] More specifically, sample collection, cfDNA extraction and sequencing. Plasma samples are obtained from healthy individuals and colorectal cancer patients, cfDNA is extracted, sequencing libraries are constructed and high-throughput methylation sequencing is performed. In this embodiment, these extraction, construction and sequencing steps all follow strict specifications, and cfDNA is efficiently extracted from plasma and sequencing libraries are constructed. The sequencing link uses existing high-throughput sequencing technologies to obtain the methylation information of cfDNA.

[0058] Collect 10 ml of whole blood samples from the subjects using a cell-free DNA collection tube. After separating the plasma by centrifugation, use the plasma DNA extraction kit provided by GENFINE company and strictly operate according to the product instruction manual to extract cfDNA. Subsequently, construct a library for the extracted cfDNA using the aforementioned patented technology. After the library construction is completed, perform RRBS sequencing. After the sequencing is completed, strict quality control needs to be carried out on the original data to screen out high-quality sequencing data for in-depth analysis.

[0059] Furthermore, the data quality control sub-module includes:

[0060] The first quality control unit is used to evaluate the quality of the initial raw sequencing data using the FastQC tool to remove low-quality bases and obtain the first quality control data;

[0061] The second quality control unit is used to remove the adapter sequences of the first quality control data using Trimmomatic or Cutadapt to obtain the second quality control data;

[0062] The third quality control unit is used to align the second quality control data to the human reference genome hg19 using the Bsmap tool to obtain the third quality control data;

[0063] The screening unit is used to obtain the reads ratio of the third quality control data aligned to the reference genome, determine whether the reads ratio meets the ratio threshold range. If so, obtain the original data of the final DNA methylation characteristics.

[0064] More specifically, use the FastQC tool to evaluate the quality of the raw sequencing data to remove low-quality bases. At the same time, use efficient tools such as Trimmomatic or Cutadapt to remove adapter sequences. Next, use the Bsmap tool to accurately align the processed reads to the human reference genome hg19. To ensure data quality, it is also necessary to calculate the reads ratio aligned to the reference genome (ensure it reaches more than 60%), the reads ratio uniquely aligned to the reference genome (also ensure it reaches more than 60%), and the reads ratio not aligned to the reference genome (ensure it is less than 10%). Only the data with qualified quality control will be used for subsequent analysis.

[0065] Furthermore, the risk prediction module 2 includes:

[0066] The feature extraction sub-module is used to extract features from the original data of the final DNA methylation characteristics to obtain a methylation feature set, a terminal motif feature set, and a breakpoint motif feature set;

[0067] The model construction sub-module is used to construct a colorectal cancer screening model set according to machine learning algorithms;

[0068] The risk prediction sub-module is used to input the methylation feature set, the terminal motif feature set, and the breakpoint motif feature set into the colorectal cancer screening model set to obtain a prediction result.

[0069] Furthermore, the feature extraction sub-module includes:

[0070] The first feature extraction unit is used to calculate the methylation status of multiple haplotypes of the original data of the final DNA methylation features to obtain a methylation feature set;

[0071] The second feature extraction unit is used to extract the first preset number of bases at the 3' end of the reads of the original data of the final DNA methylation features and calculate the distribution features to obtain a terminal motif feature set;

[0072] The third feature extraction unit is used to extract the second preset number of upstream bases and downstream bases at the 3' end of the reads of the original data of the final DNA methylation features to obtain a breakpoint motif feature set.

[0073] Specifically, the acquisition of the methylation feature set is achieved by calculating the methylation status of multiple haplotypes (covering 5 types: MM, MHL, Entropy, PDR, and CHALM) within the tumor-specific CpG island region of the sample. The identification of the tumor-specific CpG island region relies on a discovery cohort containing 106 colorectal cancer patients and healthy control individuals. By performing high-throughput methylation sequencing on the tumor tissues and corresponding normal tissue samples of the patients in this cohort, and comparing the methylation signal differences of the CpG island region between tumor and normal tissues, these specific regions are identified. This process includes: subdividing the genome into multiple CpG island regions, then systematically counting the methylation levels of various haplotypes (haplotype types include MM, MHL, Entropy, PDR, and CHALM) within each region, and further screening out the tumor-specific CpG island regions with significant differences between tumor tissues and normal tissues. The differential screening is based on the Benjamini-Hochberg corrected P value < 0.01 and the differential expression level threshold < 0.1.

[0074] By extracting the 6-base sequence at the 3' end of each read in the cfDNA high-throughput sequencing data, the distribution characteristics of the terminal motif are calculated; at the same time, the sequences of 3 base pairs upstream and downstream of the 3' end of each read are focused on to reveal the distribution of breakpoint motif features different from the terminal motif.

[0075] More specifically, for the extraction of the end motif features, a 6-base sequence at the 3'-end of each read is extracted from the cfDNA high-throughput sequencing data, and its distribution characteristics are calculated. The specific method is as follows: Align the sequencing data to the reference genome, determine the coordinates of the reads on the reference genome, take the p bases (p ∈ [4, 10]) at the 3'-end as the first base fragment, calculate its distribution ratio in the sample, and obtain the end motif features. For the extraction of the breakpoint motif features, analyze the sequence features of q base pairs (q ∈ [2, 5]) upstream and downstream of the 3'-end of the reads. After aligning the sequencing data to the reference genome, take q bases upstream and q bases downstream of the 3'-end as the second base fragment, calculate its distribution ratio in the sample, and obtain the breakpoint motif features.

[0076] Furthermore, the colorectal cancer screening model set contains 28 basic prediction models, and each of the basic prediction models is constructed based on any one of the algorithms of support vector machine, random forest, XGBoost, and CatBoost.

[0077] Specifically, the pre-trained colorectal cancer screening model is divided into three stages: discovery, training, and validation. In the discovery stage, a discovery cohort consisting of 106 colorectal cancer patients and healthy controls is used, and high-throughput methylation sequencing is performed on their tumor tissues and blood samples. Using the high-throughput methylation sequencing data of tumor tissues and corresponding normal tissue samples, the genome is divided into multiple CpG island regions, and the methylation levels of different haplotypes within each region are statistically analyzed (haplotype types include MM, MHL, Entropy, PDR, and CHALM), and tumor-specific CpG island regions are identified. Calculate the methylation levels of different haplotypes within the tumor-specific CpG island regions, which is the set of methylation features for differentiating cancer. Align the cfDNA methylation sequencing data of blood samples with the reference genome, determine the coordinates of all reads data on the reference genome, and use the p bases (p is any integer between 4 and 10) at the 3' end of the reads data on the human reference genome as the first base fragment, and calculate the distribution ratio of these fragments in the entire sample to determine the end motif characteristics; use the q bases upstream and downstream of the 3' end of the reads data on the human reference genome (q is any integer between 2 and 5) as the second base fragment, and calculate the distribution ratio of these fragments in the entire sample to determine the breakpoint motif characteristics. Determine the end motif and breakpoint motif characteristics. The criteria for differential feature screening are based on a multiple-testing corrected P-value < 0.01 and a differential expression level threshold of 0.1. Integrate the methylation features of 5 haplotypes, sequence end motif features, and breakpoint motif features to construct a comprehensive biomarker system containing seven features. These features are organized into multiple CSV files (such as the MM methylation feature screening result: Colorectum_cancer_markers_MM.csv, etc.) for subsequent analysis and application.

[0078] After the feature screening and integration are completed, we proceed to construct an early colorectal cancer screening model based on seven features. For this purpose, blood samples of another 1863 independent subjects are introduced and randomly divided into a training set (931 cases) and a test set (932 cases). The training set is used for model construction, while the test set is used to evaluate the performance of the model. Using the RRBS sequencing data in the training set, the samples are input into four different machine learning algorithms (support vector machine, random forest, XGBoost, and CatBoost). Each algorithm outputs the probability of cancer occurrence, thus forming 28 basic models, which constitute the first layer of the model. Subsequently, the prediction results of these basic models are further integrated into a logistic regression model in the second layer to optimize the final screening results. This method is called the stack ensemble learning method, which optimizes the combination method of the prediction results of different basic learners through a meta-learning algorithm, thus giving full play to the synergistic effect of multiple high-performance models and achieving more accurate classification prediction. Specifically, when training the model, the support vector machine is set with the following parameters: kernel = "linear", scale = T, probability = TRUE, type = "C-classification". The random forest is set with the following parameters: method = "rf", prox = TRUE, ntree = 500, metric = "Accuracy". XGBoost is set with the following parameters: nrounds = 1000, params = list(booster = "gbtree", objective = "binary:logistic", eval_metric = "logloss"). CatBoost is set with the following parameters: params = list(loss_function = 'Logloss', iterations = 1000). The model shows high specificity (training set: 97.50%, test set: 97.35%), high sensitivity (training set: 91.72%, test set: 93.10%), and high accuracy (training set: 95.7%, test set: 96.03%) on both the training set and the test set.

[0079] Furthermore, personalized screening recommendations and follow-up observations. According to the prediction results, personalized screening recommendations are provided for the subjects. For subjects with a predicted cancer diagnosis, colonoscopy is strongly recommended for further confirmation; for subjects with a predicted non-cancer diagnosis, combined with their basic clinical information, it is recommended to undergo colorectal cancer screening through this system once a year. At the same time, long-term follow-up observations are carried out on the subjects to monitor the accuracy of the prediction effect. For subjects whose prediction results are inconsistent with the subsequent colonoscopy results, their subsequent situations will be continuously tracked, and relevant information will be collected for subsequent research and clinical decision-making reference.

[0080] Construction and Management of Electronic Medical Record Database: It involves establishing an electronic medical record cohort database to update and record relevant information of subjects in real time. In this embodiment, while conducting long-term follow-up observations on subjects, the accuracy of their predicted results is closely monitored. For subjects predicted to have cancer, their subsequent colonoscopy test results will be tracked and compared with the predicted results. If the subsequent colonoscopy test results are consistent with the predicted results (having cancer), subsequent information of the patient will be continuously collected, including treatment plans, treatment effects, recurrence situations. If the subsequent colonoscopy test results are inconsistent with the predicted results (not having cancer), the subsequent situation of the patient will be continuously tracked to detect whether the disease occurs. For subjects predicted not to have cancer, the test results of other subsequent methods will also be tracked. If the subsequent colonoscopy test results are consistent with the predicted results (not having cancer), a health management plan will be formulated in combination with the user's previous physical examination information, including daily life guidance, follow-up observations, specialist outpatient visits, etc. If the subsequent colonoscopy test results are inconsistent with the predicted results (having cancer), subsequent information of the patient will be continuously tracked, including basic information, clinical diagnosis, symptoms, treatment plans, treatment effects, recurrence situations, etc.

[0081] This embodiment also provides an individualized management method for colorectal cancer risk prediction, including:

[0082] Collect blood samples from subjects, extract circulating free DNA, and sequence the cfDNA using reduced representation bisulfite sequencing technology to obtain the original data of the final DNA methylation characteristics;

[0083] Perform bioinformatics analysis on the original data of the final DNA methylation characteristics using a machine learning algorithm, extract the methylation feature set, terminal motif feature set, and breakpoint motif feature set, and based on the extracted feature sets, use a pre-trained prediction model to predict the colorectal cancer risk of subjects to obtain a prediction result;

[0084] Provide an individualized colorectal cancer screening plan for subjects according to the prediction results and the clinical information of the subjects;

[0085] Conduct long-term follow-up observations on subjects, monitor the diagnosis and treatment effects of colorectal cancer, and establish an electronic medical record database.

[0086] Specifically, as Figure 2 shown, an individualized management method for colorectal cancer risk prediction can also be expressed in the following steps:

[0087] Step 1: Collect peripheral blood samples from subjects, extract genomic DNA, and isolate cfDNA from it for methylation sequencing to obtain the original sequencing data of plasma cfDNA.

[0088] Step 2: Perform quality control processing on the RRBS high-throughput sequencing data of the cfDNA to obtain high-quality sequencing data for subsequent analysis.

[0089] Step 3: Based on the data after passing quality control, statistically analyze the methylation degree of tumor-specific regions to form a methylation feature set; at the same time, extract end motif features and breakpoint motif features.

[0090] Step 4: Use a pre-trained colorectal cancer screening model, combined with the multi-omics features obtained in Step 3, to predict the colorectal cancer risk of the subject.

[0091] Step 5: According to the prediction results, provide personalized screening recommendations for the subject and conduct long-term follow-up observations.

[0092] Step 6: Establish an electronic medical record cohort database, update and record the relevant information of the subject in real time, and provide support for subsequent research and clinical decision-making.

[0093] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other.

[0094] Specific examples are used in this article to elaborate on the principles and implementation methods of the present invention. The descriptions of the above embodiments are only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, based on the idea of the present invention, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. An individualized management system for colorectal cancer risk prediction, characterized in that: include: A multi-omics data processing module, used to collect blood samples from subjects, extract circulating free DNA, and sequence the cfDNA using a simplified methylation sequencing technique to obtain final raw data of DNA methylation characteristics; A risk prediction module, for performing bioinformatics analysis on the final raw data of DNA methylation features using a machine learning algorithm, extracting a methylation feature set, a terminal motif feature set, and a breakpoint motif feature set, and based on the extracted feature set, using a pre-trained prediction model to predict the subject's risk of colorectal cancer to obtain a prediction result; Personalized diagnosis and treatment guidance module, which is used to provide subjects with personalized colorectal cancer screening plans based on the predicted results and clinical information of the subjects; The follow-up observation module conducts long-term follow-up observation of subjects, monitors the diagnosis and treatment effects of colorectal cancer, and establishes an electronic medical record database.

2. The individualized management system for colorectal cancer risk prediction according to claim 1, characterized in that: The multi-omics data processing module includes: A sample processing submodule, used for collecting a peripheral blood sample of a subject and obtaining cfDNA from the peripheral blood sample of the subject; A raw data acquisition submodule, used to perform RRBS sequencing on the cfDNA to obtain initial raw sequencing data; The data quality control submodule is used to perform quality control on the initial raw sequencing data to obtain the final raw data of DNA methylation characteristics.

3. The personalized management system for colorectal cancer risk prediction according to claim 1, characterized in that: The risk prediction module includes: A feature extraction submodule, used for extracting features from the final raw data of DNA methylation features to obtain a methylation feature set, an end motif feature set, and a breakpoint motif feature set; The model building submodule is used to build a colorectal cancer screening model set according to a machine learning algorithm; The risk prediction submodule is used to input the methylation feature set, the terminal motif feature set, and the breakpoint motif feature set into the colorectal cancer screening model set to obtain a prediction result.

4. The individualized management system for colorectal cancer risk prediction according to claim 3, characterized in that: The colorectal cancer screening model set includes 28 basic prediction models, and the basic prediction models are all constructed based on any one of the algorithms including support vector machine, random forest, XGBoost and CatBoost.

5. The individualized management system for colorectal cancer risk prediction according to claim 2, characterized in that: The data quality control submodule includes: A first quality control unit, used to perform quality assessment on the initial raw sequencing data using a FastQC tool to remove low-quality bases to obtain first quality control data; A second quality control unit, used for removing the first quality control data connector sequence by using Trimmomatic or Cutadapt to obtain second quality control data; A third quality control unit is used to align the second quality control data to the human reference genome hg19 using a Bsmap tool to obtain third quality control data; A screening unit is used to obtain the reads ratio of the third quality control data aligned to the reference genome, determine whether the reads ratio meets the ratio threshold range, and if so, obtain the final raw data of the DNA methylation feature.

6. The individualized management system for colorectal cancer risk prediction according to claim 3, characterized in that: The feature extraction submodule includes: A first feature extraction unit, used for calculating the methylation states of multiple haploids of the final raw data of DNA methylation features to obtain a methylation feature set; A second feature extraction unit is used to extract a first preset number of bases at the 3' end of the reads of the final raw data of the DNA methylation feature and calculate the distribution feature to obtain a terminal motif feature set; The third feature extraction unit is used to extract a second preset number of upstream bases and downstream bases from the 3' end of the reads of the final raw data of the DNA methylation feature to obtain a breakpoint motif feature set.

7. The individualized management system for colorectal cancer risk prediction according to claim 6, characterized in that: The types of haploids include: MM, MHL, Entropy, PDR and CHALM.

8. A personalized management method for colorectal cancer risk prediction, characterized in that: include: Collecting blood samples from subjects, extracting circulating free DNA, and sequencing the cfDNA using simplified methylation sequencing technology to obtain final raw data of DNA methylation characteristics; Using a machine learning algorithm to perform bioinformatics analysis on the final raw data of the DNA methylation features, extracting a methylation feature set, a terminal motif feature set, and a breakpoint motif feature set, and based on the extracted feature set, using a pre-trained prediction model to predict the subject's risk of colorectal cancer to obtain a prediction result; Provide subjects with personalized colorectal cancer screening plans based on the prediction results and clinical information of the subjects; The subjects will be followed up for a long time to monitor the diagnosis and treatment effects of colorectal cancer, and an electronic medical record database will be established.

Images